Hydraulic system of excavator has strong nonlinearity.The control method based on model is widely used in the existing excavator control system. It is necessary to build an accurate excavator model, but the cost is too high and the control effect is poor. Therefore, an online learning control method for hydraulic excavator position was proposed based on the echo state network, and the basic online learning model was established. The model included two echo state networks,an inverse of the learning objective, and a control input based on the inverse of the learning objective. After further optimized the model, an online learning optimization model was proposed. Sinusoidal signal was used as a reference trajectory,the simulation study of the basic model and the optimization model was carried out, and the control test device of the excavator was built.Single joint motion, multi joint movement and actual excavation motion experiments were carried out respectively.The results show that by using the online learning control method, the position control accuracy of the excavator is obviously improved, and the root mean square error is reduced by more than 50%, which proves the performance and feasibility of the proposed control method